The Effects of the EU Equal‐Treatment Directive for Fixed‐Term Workers: Evidence From the UK
| Author | Andrea Salvatori |
| Published date | 01 June 2015 |
| Date | 01 June 2015 |
| DOI | http://doi.org/10.1111/bjir.12096 |
A large body of literature has documented that across Europe (and beyond), workers on temporary contracts are paid less, receive less training and report lower satisfaction than workers with similar observable characteristics on permanent contracts (Arulampalam et al. ; Booth et al. ; Brunello et al. ; Kahn ; OECD ). Such evidence fuelled a policy debate that eventually led the EU to adopt the Council Directive 1999/70/EC to prevent discrimination against fixed‐term workers. All member‐states have now transposed the Directive into national legislation. To the best of my knowledge, there has been very little research into the effects of the Directive on fixed‐term employment across Europe. This article appears to be the first large‐scale empirical study of the impact of the EU legislation on wage differentials in one of the member‐states, the UK.
The UK passed the Fixed Term Employees' (Prevention of Less Favourable Treatment) Regulations in July 2002, which then came into effect in October of the same year. The new regulations mandate that fixed‐term employees cannot be treated less favourably than comparable permanent employees in the same firm in terms of wages, benefits and training. Unequal treatment is accepted when it can be ‘objectively justified’, and employers can balance a less favourable condition against a more favourable one.
This article uses 11 years (1997–2007) of data from the Labour Force Survey (LFS) to study the effects of the new legislation on fixed‐term (FixT) workers' wages. The reform, therefore, falls in the middle of the interval studied, and the results are unlikely to be affected by the Great Recession of 2008. Previous literature has already documented the existence of a conditional wage penalty for these workers in the UK. In particular, Booth et al. () use data from the British Household Panel Survey (BHPS) between 1991 and 1997, and find that after controlling for observable characteristics the wage penalty was around 17 per cent for men and 14 per cent for women. In principle, there was therefore ample room for the new legislation to produce visible effects on the average wage differentials. However, these differentials might at least in part reflect differences across groups that researchers are not able to account for. If that was the case, the legislation could alter the market price of fixed‐term contracts, possibly leading to unintended and potentially perverse effects, for instance by making it more difficult for low‐skilled workers to obtain fixed‐term jobs due to increased cost of such contracts. As part of a wider empirical strategy to try and isolate the causal effect of the legislation on wages, I also present evidence on the possible unintended effects of the reform on the level and composition of temporary employment.
A first simple evaluation of the impact of the new legislation was conducted by Green (). He looks at the years immediately before and after 2002, and finds that the average wage of fixed‐term workers increased more than that of permanent workers. The analysis of this article attempts to address some of the potential limitations of such an approach. In the first place, a number of control variables covering an 11‐year period are included in the econometric specifications to account for possible changes in the characteristics of workers employed on different contracts. A careful and detailed descriptive analysis is certainly of interest in itself given the very limited amount of research on this topic, but this article attempts to go further and investigates the credibility of the assumptions necessary for the identification of the causal effect of the reform. In particular, the analysis is cast in a difference‐in‐difference framework, and two alternative control groups are considered, that is permanent workers (who have open‐ended employment contracts with the firm) and temporary agency workers (TAW, who are employed for a limited time by the firm through a third agency). The fundamental identification assumption is that the treated and the control groups share a common time trend, that is that in the absence of the policy intervention the wages of the two groups would have followed the same pattern. The validity of such assumption is scrutinized using the evidence on the behaviour of the relevant wage differentials before the reform. In addition, changes in the composition of unobserved characteristics must also be ruled out as they would confound the effects of the new regulations. In the absence of data that allow to tackle this issue directly, a number of checks are performed to try and uncover any indirect evidence of changes in unobservables within contract groups.
The analysis finds that fixed‐term workers of both genders suffered a conditional wage penalty of around 0.04 log points compared with permanent worker in the five years prior to the reform. For males only, there is evidence that the wage gap closed following the introduction of the equal‐treatment legislation. However, we find evidence that the closing of the gap had begun even before 2002, and that after that year the wages of agency workers, who were not subject to the new legislation, also increased. These facts cast doubts on the extent to which the disappearance of the conditional wage penalty for male fixed‐term workers can be ascribed to the equal‐treatment legislation.
The new regulations mandated equal treatment between fixed‐term workers and permanent workers, and did not apply to TAW. As will become clearer in the discussion below, the key to isolating the impact of the legislation is to identify a group of workers whose wages were behaving similarly to those of fixed‐term workers before the reform was introduced. On the one hand, permanent workers are arguably likely to meet this requirement because fixed‐term and permanent workers are both employees of the firm, whereas TAWs are employed through a third party (the agency). On the other hand, it could be argued that the market for fixed‐term jobs is more closely related to that for agency jobs because of the common temporary nature of these jobs. Therefore, this article considers both candidate groups.
The analysis begins by estimating a simple log‐linear regression model:
Consider first the possibility of general equilibrium effects. Standard economic theory suggests that the increase in the wage of fixed‐term workers relative to that of agency workers would result (a) in a decrease in the supply of and (b) in an increase in the demand for agency workers. The increase in the wages generated by such forces would be picked up by a positive γ4. To the extent that such effects take time to unfold, their empirical relevance can be assessed by considering different time intervals around 2002.
Consider now the scenario where agency workers' wages changed after 2002 as a consequence of unobservables affecting exclusively agency workers. There were no significant changes in the regulations affecting agency workers' working conditions in the UK over this period, but differences in pre‐refom trends in the wages for different contracts could explain post‐reform differences, leading to γ4 ≠ 0. Perhaps more importantly, the same concern can be raised with reference to γ2, which is meant to isolate the effect of the new legislation on the differential between fixed‐term and permanent workers. To try and purge the post‐2002 estimates from the confounding effects of differences in underlying trends, contract‐specific (quarterly) linear trends in wages can be added to equation (Angrist and Pischke ). In addition, differences in changes in wages over time could also be due to differences in the responsiveness of wages of different types of contracts to the economic cycle. Therefore, following Kugler et al. (), I also include group‐specific cyclical effects as interactions between quarterly real gross domestic product growth and the contract dummies.
Finally, factors other than the new legislation affecting both agency workers and fixed‐term workers might result in γ4 ≠ 0. Under the assumption that the effect of these unobservables on fixed‐term and agency workers is the same, one can hope to purge such confounding effects from the estimate of the policy effect by subtracting γ4 from γ2. An estimate of this difference and its standard error can be obtained by defining an additional dummy variable identifying both fixed‐term and agency workers:
To continue reading
Request your trial